5 ms·

So far I’ve used six languages to build everything from web apps to desktop applications and C++ is still by far my favourite language. It is sad that it is st

So far I’ve used six languages to build everything from web apps to desktop applications and C++ is still by far my favourite language. It is sad that it is still haunted by prejudice due to pre-C++11 era but modern C++ is easy to learn and use, and I recommend every programmer should give it a try.

pjmlp 4y ago
The problem is that many keep teaching C++ with code that a C compiler won't have any issues with. That is why Bjarne keeps doing those advocacy talks about best practices.
- ModernMech 4y ago
  The real problem as I see it is that people think the path to safe C++ is to tell everyone "you're holding it wrong" (just use modern C++), when it fact being able to hold it wrong is regarded as a feature (C++ maintains backward compatibility) by the C++ community.
  - pjmlp 4y ago
    By C expats on C++ country.
- mk89 4y ago
  Related to this, maybe: an issue I had was also the fact that online many of the help pages are really outdated. I can't forget an old accepted "much voted" answer I saw on stackoverflow which claimed that raw pointers should be owning resources, because that's how the language always did. I think the answer was dated like 2008 or so, can't find the reference anymore. That's when I started to pay a lot more attention to the published date whenever I search for help and I stumble upon a very old article. If only we could get rid of old resources written before maybe 2018-2019 and only select those that are still "modern", that'd be already a great step forward. Luckily we have this now [0]. They should invest more in tools like the GSL extension at least as a clang warning which really helps, and things like that. That's how I believe Java was also able to modernize. You can write pretty crappy Java too, but tools like findBugs or checkstyle do help, so why not! [0]: https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#r3-a-raw-pointer-a-t-is-non-owning https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines...
  - pjmlp 4y ago
    That is what using VC++ with integrated support for Core Guidelines feels like, naturally that is a minority. https://learn.microsoft.com/en-us/events/pure-virtual-cpp-2022/everything-i-learned-about-static-analysis-and-program-safety-in-cpp https://learn.microsoft.com/en-us/events/pure-virtual-cpp-20...
    - mk89 4y ago
      Yeah, I kind of envy that. Although with Clang-tidy I sometimes see GSL warnings: https://clang.llvm.org/extra/clang-tidy/ https://clang.llvm.org/extra/clang-tidy/ which is pretty cool - I wonder how up-to-date that is compared to the GSL extension. The more you invest in this kind of automated tools, the earlier you catch such bugs. Sure, sometimes it's "false positives, I know what I am doing", but you cut maybe a good > 90% of crap that comes from just writing old-style code.

pjmlp 4y ago

The problem is that many keep teaching C++ with code that a C compiler won't have any issues with. That is why Bjarne keeps doing those advocacy talks about best practices.

ModernMech 4y ago
The real problem as I see it is that people think the path to safe C++ is to tell everyone "you're holding it wrong" (just use modern C++), when it fact being able to hold it wrong is regarded as a feature (C++ maintains backward compatibility) by the C++ community.
- pjmlp 4y ago
  By C expats on C++ country.
mk89 4y ago
Related to this, maybe: an issue I had was also the fact that online many of the help pages are really outdated. I can't forget an old accepted "much voted" answer I saw on stackoverflow which claimed that raw pointers should be owning resources, because that's how the language always did. I think the answer was dated like 2008 or so, can't find the reference anymore. That's when I started to pay a lot more attention to the published date whenever I search for help and I stumble upon a very old article. If only we could get rid of old resources written before maybe 2018-2019 and only select those that are still "modern", that'd be already a great step forward. Luckily we have this now [0]. They should invest more in tools like the GSL extension at least as a clang warning which really helps, and things like that. That's how I believe Java was also able to modernize. You can write pretty crappy Java too, but tools like findBugs or checkstyle do help, so why not! [0]: https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#r3-a-raw-pointer-a-t-is-non-owning https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines...
- pjmlp 4y ago
  That is what using VC++ with integrated support for Core Guidelines feels like, naturally that is a minority. https://learn.microsoft.com/en-us/events/pure-virtual-cpp-2022/everything-i-learned-about-static-analysis-and-program-safety-in-cpp https://learn.microsoft.com/en-us/events/pure-virtual-cpp-20...
  - mk89 4y ago
    Yeah, I kind of envy that. Although with Clang-tidy I sometimes see GSL warnings: https://clang.llvm.org/extra/clang-tidy/ https://clang.llvm.org/extra/clang-tidy/ which is pretty cool - I wonder how up-to-date that is compared to the GSL extension. The more you invest in this kind of automated tools, the earlier you catch such bugs. Sure, sometimes it's "false positives, I know what I am doing", but you cut maybe a good > 90% of crap that comes from just writing old-style code.